DMM resistance measurements: 4-wire zero source

[splin]

Why do you need a 4-wire zero? The whole point of 4-wire measurements is that only the voltage across the resistor matters so why isn't it sufficient to short the sense inputs when taking a zero measurement? Is it to ensure that any common mode voltage caused by the current force circuit is also applied to the zero measurment?

I have a Datron 1081 and need to make a 4-wire zero source. Does anyone have a picture of one?

The diagrams in the manual don't make much sense to me:



What is the resistor shown? Why would it need to be shorted with a copper link for high resistance ranges but not lower ones? This seems to be the opposite of what I would expect.

[TiN]

You need a tetrajunction.
This was discussed before

[splin]

You need a tetrajunction.
This was discussed before

I'm pretty sure I don't. I did read that thread and I don't remember reading any explanation of the relevance of a tetrajunction to a 4-wire zero for a DMM. ...  Ok. I've just re-read it and am still none the wiser as to why I'd need one.

If I were to use two instruments, a constant current source and a voltmeter to measure a resistor then I would zero the voltmeter by shorting its terminals with some thin copper wire (big copper bars, links etc. are thermally terrible for this and quite unnecessary given there should be almost no current flowing through the short).

The voltmeter, being isolated and floating wrt the current source, means that it would be irrelevant if the current source were connected to one or other of the voltmeter terminals or not.

In the case of the DMM however I assume that the zero source sense and force interconnections are related to common mode voltage errors and reflect a limitation of combining the two instruments into one for convenience?

[Dr. Frank]

Why do you need a 4-wire zero? The whole point of 4-wire measurements is that only the voltage across the resistor matters so why isn't it sufficient to short the sense inputs when taking a zero measurement? Is it to ensure that any common mode voltage caused by the current force circuit is also applied to the zero measurment?

I have a Datron 1081 and need to make a 4-wire zero source. Does anyone have a picture of one?

The diagrams in the manual don't make much sense to me:


What is the resistor shown? Why would it need to be shorted with a copper link for high resistance ranges but not lower ones? This seems to be the opposite of what I would expect.

That technique of 4W zero obviously mimics the real True Ohm function of the 1281 / 8508A, that is the Offset Compensation of the 3458A.
The purpose is to measure and cancel all effective e.m.f. thermo couples, which are in series inside the instrument, and at the DUT.

For high resistance DUTs, you get more noise, and you might get bias current effects, so that offset compensation might not work well.
That might be the reason, why they propose using an additional short.
Due to these problems, instruments with built in offset compensation, limited that technique to the 20kOhm, or to the 100kohm range.

Anyhow, this short at the DUT for high resistance makes no sense to me, at least it is quite an unclear situation, which e.m.f. is really measured, that of the DUT to the measuring cables, or of the short to the cables.

For low resistance values, neither noise, nor bias current are a problem, so the offset reading can be done using the DUT only.


It's more precise to connect the I+ to the I- source node. If you would simply short I+ and I- , the test current would create a voltage drop across the zeroing cable, introducing a considerable error.
 
At 100 Ohm for DUT, and 100 mOhm for the test cable, that might be a 0.1% error, for 10k DUT, it's still 10ppm, well observable on 6 1/2 digit DMMs.

Frank

[dl1640]

The Hi lead and I+ lead are paired, so are the Lo lead and I- lead.

Why it is configured like that? (Normally we have Hi and Lo as one pair)

[e61_phil]

That technique of 4W zero obviously mimics the real True Ohm function of the 1281 / 8508A, that is the Offset Compensation of the 3458A.
The purpose is to measure and cancel all effective e.m.f. thermo couples, which are in series inside the instrument, and at the DUT.

The purpose on the 8508A (and I think also on the 1281) is the same, but True Ohms is reversing the current instead of switching it off, like the 3458A and 3456A (and others) do. There is no meaurement at 0V, but on reversed polarity (which is also reversed for the ADC in the 8508A). The effect is almost the same, but it has the advantage of constant power at the resistor and maybe the disadvantage of a doubled voltage swing.

One can run a Zero compensation on the 8508A even in True Ohms, but I have no idea why that should give a benefit. All these techniques with two different currents (no matter if one of these is 0A, -i or any other level) are able to compensate for an offset.

The Hi lead and I+ lead are paired, so are the Lo lead and I- lead.

Why it is configured like that? (Normally we have Hi and Lo as one pair)

If you have leakage between the pairs it doesn't matter in this configuration, because there is no voltage across HI / I+ and LO / I-

It's more precise to connect the I+ to the I- source node. If you would simply short I+ and I- , the test current would create a voltage drop across the zeroing cable, introducing a considerable error.

But, if you disconnect HI/LO and I+/I- completly, you have to take care if the circuit can handle this, if the poentials drift away from each other.


To your question at the beginning: I would say it is exactly as Frank said: For low resistors you can use the DUT itself for the short. For higher resistors the input current of the meter creates some voltage over the DUT, therefore the short. For me the manual makes sense.

[David Hess]

Why do you need a 4-wire zero? The whole point of 4-wire measurements is that only the voltage across the resistor matters so why isn't it sufficient to short the sense inputs when taking a zero measurement?

The image you included says exactly why, "... which ensure that thermal emf effects are eliminated."

4 wire measurements are used where lead resistance is significant which happens at low voltages.  Low voltages are also where thermal emf effects become significant so to take best advantage of a 4 wire measurement, the thermal emf effects must be removed.

Another way to do this is to reverse the force and sense leads, make a second measurement, and average the two measurements.  Some meters do this automatically.  Reversing the current inverts the voltage but the thermal emf errors remain the same so are removed.

[dl1640]

The True Ohm on 1271 is probably like the OComp on 3458, current source is on and off, it has no reverse current source.

[splin]

That technique of 4W zero obviously mimics the real True Ohm function of the 1281 / 8508A, that is the Offset Compensation of the 3458A.
The purpose is to measure and cancel all effective e.m.f. thermo couples, which are in series inside the instrument, and at the DUT.

Yes, I realise that, but I didn't understand exactly what the diagram was trying to show - I assumed it was referring to specially made 4-wire zero jigs used to set the zero level. It didn't make sense though as there would be no reason to not use a copper shorting strap whatever the range being used.

For high resistance DUTs, you get more noise, and you might get bias current effects, so that offset compensation might not work well.
That might be the reason, why they propose using an additional short.

Ok, from the above and e61_phil's subsequent reply, I see they were referring to using the DUT itself which makes much more sense. I believe Keithley (or Fluke?) actually supply 4-wire zero devices which plug into the meter terminals, so that added to my uncertainty.

Due to these problems, instruments with built in offset compensation, limited that technique to the 20kOhm, or to the 100kohm range.

Anyhow, this short at the DUT for high resistance makes no sense to me, at least it is quite an unclear situation, which e.m.f. is really measured, that of the DUT to the measuring cables, or of the short to the cables.

For low resistance values, neither noise, nor bias current are a problem, so the offset reading can be done using the DUT only.


It's more precise to connect the I+ to the I- source node. If you would simply short I+ and I- , the test current would create a voltage drop across the zeroing cable, introducing a considerable error.

Not sure what you mean - when you say source node do you mean the meter terminals? But either way the test current would not be flowing through the sense leads so there should be no error - that is the point of 4-wire after all. You want to measure zero with the test leads/DUT configuration as close as possible to when making the actual resistance measurement.

Anyhow, looking at the way Datron implemented ohms measurments I can now see why the zero is set up the way they suggest as the meter doesn't directly measure the voltage across the sense leads - rather it measures Sense Hi wrt ground and adjusts Force+, which in conjunction with Force-, servos Sense Lo to be equal to ground:



In other words, Force+ and Force- must to be connected to Sense Lo to allow the meter to force Lo to be equal to ground.

Interestingly I see that Datron changed the 4-wire zero arrangement slightly for the 1281:



Presumably this was to eliminate the error produced by the meter's input bias current flowing through the DUT resistance:






In the latter case the meter's bias current does not flow through Rx (apart from the voltage follower's bias current) but the DUT to Sense Hi and Lo thermal EMFs will still be measured at Sense Hi relative to ground.

[splin]

Why do you need a 4-wire zero? The whole point of 4-wire measurements is that only the voltage across the resistor matters so why isn't it sufficient to short the sense inputs when taking a zero measurement?

The image you included says exactly why, "... which ensure that thermal emf effects are eliminated."

4 wire measurements are used where lead resistance is significant which happens at low voltages.

Low voltages are also where thermal emf effects become significant so to take best advantage of a 4 wire measurement, the thermal emf effects must be removed.

Or more accurately at high test currents needed to develop the required test voltage. The 3458A uses 1V for the 10k range and below and 5V for 100K and above. Thermal EMFs are not much less a significant source of error when measuring 5V than 1V but presumably the higher noise and leakage / bias current errors when measuring higher resistance values mean that the thermal EMFs are swamped and so there is no point in performing Offset compensation - especially given the disadvantage of slower readings, settling times etc.

Another way to do this is to reverse the force and sense leads, make a second measurement, and average the two measurements.  Some meters do this automatically.  Reversing the current inverts the voltage but the thermal emf errors remain the same so are removed.

Yes I appreciate that - what I couldn't understand was why you needed a 4-wire zero rather 2-wire zero where the thermal offset voltages are measured across the resistor with no current flowing - and as there is no current, I couldn't see why you needed the force leads connected at all. However, as I put in my last post, it is clear that it is necessary for the Datron meters.

[David Hess]

Yes, I realise that, but I didn't understand exactly what the diagram was trying to show - I assumed it was referring to specially made 4-wire zero jigs used to set the zero level. It didn't make sense though as there would be no reason to not use a copper shorting strap whatever the range being used.

I think what they were getting at is that you must do the zero measurement with the assembly you intend to use to make the resistance measurement because if you change it, then the thermal emf voltages will be different.  So there is no separate "zero jig".

[e61_phil]

I wasn't aware of the fact, that the 1081 uses the same Ohms circuit like the 1281 and 8508A. This circuit has a couple of advantages above the 3458A. The LO input is forced to the GND potential which is the same as the Ohms Guard potential. The I+/HI side is forced by the current sourcing opamp. That means: If your cables are leaky the leakage from the HI/I+ side doesn't matter, because it is a low impedance voltage source. The LO/I- inputs are on the same potential as your Ohms Guard. That means you can apply the same potential on both side of the leakage resistance and no current will flow through that leak. That makes it even possible to measure resistors within a circuit and helps to reduce leakage and noise a lot. The 8508A is in some Ohms ranges more than an order of magnitude more stable than the 3458A.

[Dr. Frank]

It's more precise to connect the I+ to the I- source node. If you would simply short I+ and I- , the test current would create a voltage drop across the zeroing cable, introducing a considerable error.

Not sure what you mean - when you say source node do you mean the meter terminals? But either way the test current would not be flowing through the sense leads so there should be no error - that is the point of 4-wire after all.

I was answering to: 'so why isn't it sufficient to short the sense inputs when taking a zero measurement? '

I interpreted your idea like this, R_sh. being the resistance of your external short:




There's no essential difference, where you actually apply this short, i.e. directly at the sense jacks, or at the DUTs sense connections; the way I sketched it, you would unambiguously measure the e.m.f.

The e.m.f. compensated reading would compute like this:

U₁ = I₊ * R_x + U_emf
U₂ ~ I₊ * R_sh. + U_emf

U_comp = U₁ - U₂ = I₊ * (R_x - R_sh.)

So the DUT would be measured low by the amount of the shorting resistance, which may be between 10 to 100 mOhm.

Frank

[Kleinstein]

I wasn't aware of the fact, that the 1081 uses the same Ohms circuit like the 1281 and 8508A. This circuit has a couple of advantages above the 3458A. The LO input is forced to the GND potential which is the same as the Ohms Guard potential. The I+/HI side is forced by the current sourcing opamp. That means: If your cables are leaky the leakage from the HI/I+ side doesn't matter, because it is a low impedance voltage source. The LO/I- inputs are on the same potential as your Ohms Guard. That means you can apply the same potential on both side of the leakage resistance and no current will flow through that leak. That makes it even possible to measure resistors within a circuit and helps to reduce leakage and noise a lot. The 8508A is in some Ohms ranges more than an order of magnitude more stable than the 3458A.

The different Ohms circuit is not that much better:
The roles of the high and low side are changed - leakage at the high side does not matter, but leakage and bias at the low side becomes the problem now. Much of the leakage advantage is having a guard at the low side, that the 3458 and many other meters lack - they would need a guard at the high side. One the downside the extra OP at the low side can add to the errors and a possibly needed extra relay to bypass this could add errors  to voltage measurement.

One advantage of the circuit is that internal there is more headroom on the negative side for the current source and in some forms one could directly measure the current from the low side with shunt voltages near ground level.

The not so stable Ohms ranges in the 3458 are more like due to not so good parts used and having a single current source circuit used for all ranges. Ideally it takes more than just switching the resistors.

[e61_phil]

But on the 3458A there is no driven guard and, if added, it will add exactly that input current which is added by the low follower in the Datron circuit.
I don't understand the argument with the opamp and voltage meausurement swichting. Normally, the current is sourced through the "voltage" input jacks, but the follower is connected to the SENSE inputs. Therefore, it shouldn't matter for DCV measurements? Should it?

And it is also much more convient to have a guard at 0V and not at 200V.

Another benefit of the Daton circuit is, that the current source sees always a constant voltage. It operates against 0V no matter what DUT is connected. This ensures a constant current over the whole range and you only need to ensure a linear voltage measurement to have a linear resistance measurement. On circuits with the current source on top, you have to make sure, that the current stays constant with different burden voltages.
Same argument for the follower. This opamp is always operated on constant input voltages and doesn't move with the burden, which can lead to different input currents.

[dl1640]

Does anyone have access to 1271 service handbook? Or is there a pdf of it available?